Skip to Content

AR Drone Identification and Navigation Control at CVG-UPM

Publication
Congresses name: 

XXXIII Jornadas Nacionales de Automática, 2012

Place: 

Vigo, Spain

Date: 
September 5-7, 2012

This article presents the proposal of the Computer Vision Group to the first phase of the international competition "Concurso de Ingeniería de Control 2012, Control Autónomo del seguimiento de trayectorias de un vehículo cuatrirrotor". This phase consists mainly of two parts: identifying a model and designing a trajectory controller for the AR Drone quadrotor.   For the identification task, two models are proposed: a simplified model that captures only the main dynamics of the quadrotor, and a second model based on the physical laws underlying the AR Drone behavior. The trajectory controller design is based on the simplified model, whereas the physical model is used to tune the controller to attain a certain level of robust stability to model uncertainties. The controller design is simplified by the hypothesis that accurate positions sensors will be available to implement a feedback controller.

Fig1. Rigid Body model of the quadrotor and proposed controller architecture
 

The control scheme presented in this article achieved the best controller score in phase 1 of the CEA 2012 competition.  Moreover, this controller obtained the best score regarding average distance error, 0.10 m, and maximum distance error, 0.62 m, to the desired trajectory;  and also the least roundtrip time with an average speed of 1.96 m/s. The model identification obtained the second best identification score. The team attained the second position in the first phase of the competition.
 
 
Fig2. Example of speed planning attained by the proposed architecture
 
The teams from the group, an undergraduate students team tutorized by J. Pestana, and the CVG group team, attained the third and fifth positions, respectively, in the final (experimental) phase of the IC CEA 2012 competition.